Improving Emergency Medical Dispatching with Emphasis on Mass-Casualty Incidents

Transcription

1 Improving Emergency Medical Dispatching with Emphasis on Mass-Casualty Incidents, Harald Burgsteiner, Martin Bernroider, Günter Kiechle, Maria Obermayer

2 Dispatcher s Challenges Dispatching ambulances is demanding Planning about 30 to 90 minutes in advance

3 Constraints & Assumptions Pickup/delivery Separation of emergencies/routine transports 7 different ambulance types Application of time windows Multiple depots Combination of routine transports due to ambulance capacities 4 different transport types

4 Objective Function k K Ambulance k out of the set of all ambulances (i, j) A Edge between the locations i and j i N Location i out of the set of all request h P r Routine transport h out of the set of all routine locations transports w i,k Normal waiting time w i,k Excessive waiting time Beginning of the service at location i by d i,j Distance between the locations i and j ambulance k a i Appointment at location i x i,j,k 1 if edge (i, j) is traversed by ambulance k B i,k w i,k w i,k min α d i,j x i,j,k k K (i,j) A = max(0, B i,k l i ) = max(0, B i,k a i w i,k + β w i,k i N ) + γ w i,k + δ z h r h P r l i α β Point in time marking the boundary between normal and excessive waiting time Weight factor of the minimum distance objective Weight factor of the minimum waiting time objective z h r γ δ 1 if routine transport h could not be included in the solution Weight factor of the excessive waiting time Weight factor of the number of unscheduled routine transports

5 Adaptive Large Neighborhood Search Searching different areas of the solution space Application of different destroy/repair heuristics Random removal Greedy insertion Worst removal Regret-2 heuristic Automatic weight adjustment due to solution quality Acceptance of worse solutions for diversification Simulated annealing framework

6 Evaluation (I) Instance Depots #Vehicles #Requests f_1_8_ f_2_8_74 1, f_2_8_89 1, f_4_8_74 1, 4, 6, f_8_8_81 all 8 81 f_8_27_272 all Salzburg Stadt 2 Hallein 3 Saalfelden 4 Zell/See 5 Gastein 6 Tamsweg 7 Bischofshofen 8 - Lamprechtshausen

7 Evaluation (II) Qualitative evaluation Reference Start solution 1 min 5 min 50,000 it. Instance Cost b Cost n Cost b Cost n Cost b Cost n Cost b Cost n Cost b Cost n t [h] f_1_8_ % -8.1% -28.4% -13.1% -31.5% -16.9% -36.8% -23.3% 71 f_2_8_ % 8.6% -9.0% -9.0% -11.0% -11.0% -13.9% -13.9% 42 f_2_8_ % 1.5% -36.0% -14.1% -38.9% -18.0% -44.1% -25.0% 58 f_4_8_74 1, % 70.8% 8.7% 44.9% 0.0% 33.3% -8.0% -22.7% 20 f_8_8_ % 8.9% -2.5% 3.9% -4.9% 1.4% -8.8% -2.8% 27 f_8_27_272 3,171 2, % 15.9% -12.3% 10.7% -15.4% 6.8% -26.6% -7.3% 127

8 Evaluation (III) Performance evaluation Data in [s] #removed ambulances Avg. Instance f_1_8_ f_2_8_ f_2_8_ f_4_8_ f_8_8_ f_8_27_ (14.97*) avg * average value of the first 4 columns

9 Conclusions More and different destroy/repair heuristics Include rostering of personnel, ambulance coverage, weather data, traffic status Connection of optimization system to the computer aided dispatch system

10 Thank you for your attention!, MSc BSc